Breaking

SpaceXAI Releases Grok 4.6, Reaching No. 7 in Code Arena WebDev

August 12, 2026 at 11:47 EDT

Foundation Models
Software Dev & Coding
AI Agents

SpaceXAI released Grok 4.6 on August 12, positioning its new flagship model for long-running agents, coding, knowledge work, and ambitious interactive projects. At launch, Grok 4.6 High entered Code Arena WebDev at No. 7 with 1,618 points—65 points above Grok 4.5 High at No. 13.

Aug 12, 2026 · SpaceXAI

Grok 4.6 closes the WebDev gap

The new flagship enters Code Arena WebDev at No. 7, gaining 65 points over Grok 4.5 and landing within nine points of two frontier rivals.

Proportional comparison

Code Arena WebDev scores

All columns use the same point scale and proportional height.

1,553

Grok 4.5

1,618

Grok 4.6

1,622

GPT-5.6 Sol

1,627

Claude Fable 5

Grok 4.6 gains 65 points and moves from No. 13 to No. 7.

500K

Context window

$2 / $6

Input and output per 1M tokens

2×

Included usage during week one

AA Intelligence Index

Where it advances

Stronger agentic work

The model leads the listed field on GDPVal-AA v2, improves every shown result over Grok 4.5, and targets longer projects with more self-testing.

What remains open

Mixed evidence

Its 26% Terminal-Bench result trails the two listed frontier rivals, while the new WebDev rank can still move as confidence intervals tighten.

The strategic bet

Sustain useful work across many steps

Grok 4.6 is aimed at research, codebase work, analysis, and polished interactive applications—not only one-shot answers.

Training

Longer supplemental run plus agentic RL

Availability

API, Cursor, Grok Build, and partners

Next test

Independent long-run production use

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.